synthetic-live-session-replay

qmap can already render static session topologies well, and c23 is proving that custom models can inhabit those scenes. What is still unresolved is the temporal layer: how actors appear over time, how flows accumulate, how phase changes are communicated, and how much layout churn is acceptable when a session unfolds live.

If we wait for real session feeds before solving those problems, we will end up debugging data contracts, playback controls, and visual behavior all at once. A synthetic replay cycle lets us rehearse that system deliberately, with deterministic input and repeatable outcomes.

This cycle is adjacent to:

• c21.ai-session-map, which established the semantic AI-session topology
• c22.runtime-session-map, which established the reduced runtime-session topology
• c23.custom-obj-models, which uses runtime topology as a conservative model-integration sandbox

One runnable replay path that:

• plays a synthetic session over 10 seconds
• shows a clear progression from init -> research -> edit -> finalize
• emits session-shaped qmap data during playback
• supports basic playback controls and deterministic restart/seek behavior
• identifies the thinnest reusable seam for future real-data adapters

• cycles/c24.synthetic-live-session-replay/ — cycle metadata and planning docs
• dev/experiments/ — an experiment surface that can host replay playback controls and qmap output
• experiment-local replay helpers — scenario spec, timeline compiler, playback source, and adapters

Optional if clearly warranted:

• a narrow reusable playback utility in src/ if the experiment proves the seam is stable

• Cycle metadata
  • cycles/c24.synthetic-live-session-replay/CYCLE.md
  • cycles/c24.synthetic-live-session-replay/initial-prompt.md
  • cycles/c24.synthetic-live-session-replay/PLAN.md
• Runnable experiment support
  • scenario spec(s)
  • timeline compiler
  • playback source
  • minimal playback controls

• Full generalization for every future session type
• Solving final production UX for transport controls
• Replacing all existing static fixtures
• Streaming real back-end data in this cycle
• Production-grade persistence, recording, or export features

The first replay should tell one concrete story:

• 0ms - 1000ms: agent initializes
• 1000ms - 4500ms: research phase touches 24 HTTP URLs
• 4500ms - 9000ms: edit phase uses 4 MCP servers and touches 6 files
• 9000ms - 10_000ms: finalize and settle into a readable end state

The canonical topology should preserve the richer agent-session concepts:

• User
• AI Agent
• MCP Server
• File
• External Host

That keeps phase semantics and MCP usage visible instead of flattening everything into a generic runtime starburst.

The replay path should be split into three layers:

1. Scenario spec
   • human-authored story and actor roster
2. Timeline compiler
   • deterministic timed events and checkpoints
3. Playback source
   • play/pause/seek/restart and session-shaped output

The view should consume normal qmap session data and remain mostly agnostic to whether its source is static, replayed, or live.

1. What is the smallest replay contract that still preserves meaningful phase semantics?
2. How do we reveal actors over time without causing unacceptable layout churn?
3. Should playback operate on timed events, precomputed snapshots, or both?
4. What playback controls are sufficient for iteration without overbuilding UI?
5. Can one replay transport serve both semantic agent sessions and future reduced runtime sessions through adapters?
6. Does any of this need promotion into src/, or should it remain experiment-local for now?

• A deterministic 10-second replay runs end-to-end in qmap
• Playback can be paused, restarted, and scrubbed without desynchronizing state
• Phase changes are legible at a glance
• Actor and flow accumulation feels live without overwhelming the scene
• The cycle ends with a concrete recommendation for the reusable seam, if any

Keep this replay system experiment-local first. If the compiler/playback boundary stabilizes and real-data adapters fit cleanly behind it, then promote only that narrow seam.